How many emails do you turn down daily?
In 2018 the average open rate for e-commerce was 15%. That means that out of 100 emails 85 were simply not opened.
The profit of your email campaigns strongly depends on the email open rate. If nobody opens your email, no new revenue from email marketing will be generated.
But why do we open emails and ignore others?
Researches shows that 50% of people decide to open an email based on the subject line.
So, what makes a great subject line?
Should you include emojis? Punctuation symbols? What is the optimal length? Is it better to send your email on Tuesday morning rather than Friday afternoon? All these combined factors make it difficult to figure out what matters. That’s when Artificial Intelligence makes the sparks fly.
As Brian Kardon said “AI won’t write a headline or produce a quality video but it will tell you which campaign works best with which audience segment in realtime. There are thousands of variables changing every second, and humans can’t process all that information”
At Machine Labs we are data-driven. We combined probabilistic models and machine learning algorithms to analyzed thousands of emails and identified patterns of what makes a successful email marketing campaign.
Our results show that subject lines of about 40 characters with at least one punctuation symbol or emoji tend to perform better. Sending the email on a working day also increases the open rate.
But, what matters the most are the first four keywords of the subject line. The other words have almost no effect on the open rate.
DMR reports show that each day, the average office worker receives about 120 emails.
Your email will probably be evaluated extremely quickly, with most people deciding whether it justifies consideration or not within two seconds.
That’s why our subject line scoring takes into account the position of each word in the subject line. Words placed at the beginning are considered stronger words than those at the end. In this way, we were able to predict the open rate of emails within a 1-2% error margin.